基于运动轨迹的手势识别算法研究

  • 格式:pdf
  • 大小:2.69 MB
  • 文档页数:60

哈尔滨工业大学工学硕士学位论文

- I - 摘 要

近些年,物联网技术快速发展,人机交互应用越来越普遍,而手势作为

人机交互的重要一环,在人们生活中扮演了重要角色。目前,基于MEMS惯

性传感器的手势识别方法,都需要分析每个手势的加速度和角速度的特征值

或者相应的规律,然后根据这些特征值或者规律来识别手势,此方法需要分

析每个手势的运动细节,操作者需要根据设计者事先定好的手势进行操作,

这样会对用户手势操作造成不便。针对上述情况,本文提出了一种基于运动

轨迹的手势识别方法,通过追踪手势的运动轨迹来识别手势,可以避免对每

个手势进行运动学分析,只需分析手势运动轨迹特征,方便用户手势操作,

同时也增加了手势识别的多样性。

在研究了大量的相关资料的基础上,本文从基本原理出发,介绍了运动

轨迹追踪的基本原理和手势识别方法,分析了MEMS惯性传感器误差来源,

并建立了加速度和角速度的系统误差模型和随机误差模型。在运动轨迹追踪

理论基础上,设计了数据自动截断算法以减少冗余数据,设计了四元数误差

处理算法来提高转移矩阵的精度,此外由于重力加速度对运动轨迹计算的影

响,设计了重力加速度补偿算法以消除其影响,之后针对积分计算的速度和

位移存在误差问题,提出了速度位移重建算法。在手势识别理论基础上,结

合实际手势运动轨迹,设计了运动轨迹的特征值和手势分类器以及特征值的

提取算法。最后,使用Matlab对运动轨迹的手势识别算法进行仿真,并给出

了实验结果,该实验结果表明本系统的手势识别率高于90%,达到预期效果。

关键词:运动轨迹;MEMS惯性传感器;手势识别;特征值;分类器

哈尔滨工业大学工学硕士学位论文

- II - Abstract

In recent years, Internet of Things technology has been rapidly developed. Human

computer interaction is more and more common. In addition, the gesture is as an

important part of human computer interaction and plays an important role in people's

lives. However, at present, MEMS inertial sensor-based gesture recognition methods

need to analyze the eigenvalues or corresponding laws of acceleration and angular

velocity of each gesture. Then according to the above rules or eigenvalues, the system is

able to identify the corresponding gesture. The kind of method demands to analyze the

movement details of each gesture, and the operator needs to operate corresponding

gesture on the basis of the designer's predetermined gestures, This will lead to such a

result that the user needs to operate the corresponding gestures inconveniently. In view

of the above situation, this paper presented a gesture recognition method based on

motion trajectory. By tracking the motion trajectory of the gesture, the paper was able to

avoid the kinematics analysis of each gesture and just needs to analyze the gesture

trajectory characteristics. This article aims not only to facilitate user-defined gesture

operations, but also increased the diversity of gesture recognition.

After knowing a lot of documents, this paper introduced the basic principles of

motion tracking and the method of gesture recognition, and has analyzed the sources of

error of MEMS inertial sensors, and has established the acceleration and angular

velocity of the system error model and the random error model. On the one hand, based

on the trajectory tracking theory, this paper has designed an automatic data truncation

algorithm to reduce the redundant data. The quaternion error processing algorithm has

been designed to improve the accuracy of the transfer matrix. In addition, due to the

influence of gravitational acceleration on trajectory calculation, a gravitational

acceleration compensation algorithm has been designed to eliminate its influence. After

that, the error of the velocity and displacement of the integral calculation exists, and the

speed displacement reconstruction algorithm has been proposed. On the other hand,

based on the theory of gesture recognition, the paper has combined with the trajectory

of the actual gesture, and the eigenvalue and gesture classifier of trajectory and the

algorithm of extracting eigenvalue have been designed. Finally, the algorithm of gesture

recognition of motion trajectory has been simulated by Matlab. Finally the experimental

results was shown graphically. And the result demonstrates that the gesture recognition

rate of the system is higher than 90%, to achieve the desired effect.

Keywords: trajectory, MEMS inertial sensor, gesture recognition, eigenvalues,

classifier 哈尔滨工业大学工学硕士学位论文

- III - 目 录

摘 要 ................................................................................................................. I

ABSTRACT ....................................................................................................... II

第1章 绪 论 .................................................................................................. 1

1.1 课题研究背景及意义 .................................................................... 1

1.2 国内外该方向的研究现状与分析 .................................................... 2

1.2.1 运动轨迹检测研究现状 .................................................................... 2

1.2.2 手势识别研究现状 ............................................................................ 3

1.3 本文主要研究内容 ........................................................................ 5

1.4 本文结构 ..................................................................................... 5

第2章 运动轨迹的手势识别相关理论 ............................................................ 7

2.1 基于MEMS传感器的运动轨迹计算理论 ........................................ 7

2.1.1 滤波算法 ........................................................................................... 7